1256424 - Get rid of ThreadsafeAutoSafeJSContext

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Boris Zbarsky [:bzbarsky]]

It's pretty weird in that it has totally different guarantees on workers and mainthread in terms of what compartments it enters and whether it puts an AutoJSAPI on the stack.... In practice, it has two consumers, one of which should be using AutoJSAPI and one of which just wants rooting.  Let's kill this thing.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Er, I meant ThreadsafeAutoSafeJSContext.  ThreadsafeAutoJSContext will be more complicated.  ;)

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: Get rid of ThreadsafeAutoSafeJSContext

Comment 3

[Bobby Holley (:bholley)]

Comment on attachment 8730365 [details] [diff] [review]
Get rid of ThreadsafeAutoSafeJSContext

Review of attachment 8730365 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/base/DOMRequest.cpp
@@ +302,3 @@
>                         const JS::Value& aResult) :
>      mReq(aRequest),
> +    mResult(CycleCollectedJSRuntime::Get()->Runtime(), aResult)

RootingCxForThread would also be fine here. Doesn't really matter.

::: dom/promise/Promise.cpp
@@ +1763,5 @@
>      MOZ_ASSERT(mCountdown > 0);
>  
> +    AutoJSAPI jsapi;
> +    if (!jsapi.Init(mValues)) {
> +      // Now what?

We could make the object-valued init infallible, though that would mean renaming it. Either way, this comment should probably go.

Comment 4

[Boris Zbarsky [:bzbarsky]]

> We could make the object-valued init infallible

Are we sure we can't end up with the nsIGlobalObject missing its pointer back to the global JSObject* even while JSObjects from that global are reachable?

> RootingCxForThread would also be fine here.

Er, right.  I did that in the caller first, then realized I can just push it here.  I'll change to RootingCxForThread.

Comment 5

[Bobby Holley (:bholley)]

(In reply to Boris Zbarsky [:bz] from comment #4)
> > We could make the object-valued init infallible
> 
> Are we sure we can't end up with the nsIGlobalObject missing its pointer
> back to the global JSObject* even while JSObjects from that global are
> reachable?

I don't know.

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/21e5497ebb72

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a5365af3f8c6

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/21e5497ebb72
https://hg.mozilla.org/mozilla-central/rev/a5365af3f8c6